Conventionally, there has been used a computer system in which a plurality of computer devices are connected through a network based on the concept of network & computing. In this type of system, one of the computer devices collects logs concerning occurrence of errors from the other computer devices, and the errors are handled based on the logs. In such a log-collection system, it is desirable that only the most possible important logs are collected from the viewpoints of utilization efficiency of resources.
The conventional type of information collection system comprises a plurality of local machines, each of which is placed in a user side, and a remote machine that collects logs concerning occurrence of errors from the plurality of local machines via a network. In this information collection system, a local machine generates a log caused by occurrence of an error, and stores the log on memory. When the memory becomes full, the local machine notifies the remote machine via the network that the memory is full.
Accordingly, the remote machine collects the logs from the relevant local machine via the network and stores the logs on memory. From then on, the remote machine performs collection and storage of logs each time the remote machine receives notification from respective local machines. Further, when generating a log after the memory becomes full, the local machine overwrites the memory with the relevant log. Accordingly, the old log stored on the memory is deleted.
As explained above, in the conventional type of information collection system, since the remote machine collects logs regardless of their priorities or importance, there may come up such circumstances that a large amount of logs with low priorities may be collected whereas logs with high priorities may hardly be collected.
Therefore, under such circumstances, there have been extremely serious problems that error handling based on the logs with high priorities may be delayed or error handling can not be performed if the worst happens. Further, in the conventional type of information collection system, there is an upper limit to memory capacity of a remote machine and a local machine. Therefore, there has been a problem that utilization efficiency of memory resources is degraded when a large amount of logs with low priorities is stored on such memory.
Further, in the conventional type of information collection system, an old log is deleted by being overwritten with a new log at the point of time the memory becomes full. Therefore, if the importance of the old log is extremely high (e.g., the log caused by occurrence of a critical error), the system can do nothing about the error handling, which has a profound effect on the system.